TY - JOUR T1 - Predicting Infection-related Consultations on Intensive Care Units - Development of a Machine Learning Prediction Model JF - medRxiv DO - 10.1101/2021.03.31.21254530 SP - 2021.03.31.21254530 AU - Christian F. Luz AU - Dimitrios Soudis AU - Maurits H. Renes AU - Leslie R. Zwerwer AU - Nicoletta Giudice AU - Corinna Glasner AU - Maarten W. N. Nijsten AU - Bhanu Sinha Y1 - 2021/01/01 UR - http://medrxiv.org/content/early/2021/04/06/2021.03.31.21254530.abstract N2 - Objectives Infection-related consultations on intensive care units (ICU) build an important cornerstone in the care for critically ill patients with (suspected) infections. The positive impact of consultations on quality of care and clinical outcome has previously been demonstrated. However, timing is essential and to date consultations are typically event-triggered and reactive. Here, we investigate a proactive approach by predicting infection-related consultations using machine learning models and routine electronic health records (EHR).Methods We used data from a mixed ICU at a large academic tertiary care hospital including 9684 admissions. EHR data comprised demographics, laboratory results, point-of-care tests, vital signs, line placements, and prescriptions. Consultations were performed by clinical microbiologists. The predicted target outcome (occurrence of a consultation) was modelled using random forest (RF), gradient boosting machines (RF), and long short-term memory neural networks (LSTM).Results Overall, 7.8 % of all admission received a consultation. Time-sensitive modelling approaches and increasing numbers of patient features (parameters) performed better than static approaches in predicting infection-related consultations at the ICU. Splitting a patient admission into eight-hour intervals and using LSTM resulted in the accurate prediction of consultations up to eight hours in advance with an area under the receiver operator curve of 0.921 and an area under precision recall curve of 0.673.Conclusion We could successfully predict of infection-related consultations on an ICU up to eight hours in advance, even without using classical triggers, such as (interim) microbiology reports. Predicting this key event can potentially streamline ICU and consultant workflows and improve care and outcome for critically ill patients with (suspected) infections.Competing Interest StatementThe authors have declared no competing interest.Funding StatementThis study was part of a project funded by the European Commission Horizon 2020 Framework Marie Skłodowska-Curie Actions (grant agreement number: 713660-PRONKJEWAIL-H2020-MSCA-COFUND-2015).Author DeclarationsI confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained.YesThe details of the IRB/oversight body that provided approval or exemption for the research described are given below:Ethical approval was obtained from the institutional review board of the University Medical Center Groningen and informed consent was waived due to the retrospective observational nature of the study (METc 2018/081).All necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived.YesI understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance).YesI have followed all appropriate research reporting guidelines and uploaded the relevant EQUATOR Network research reporting checklist(s) and other pertinent material as supplementary files, if applicable.YesThe data is registered in the Groningen Data Catalogue (https://groningendatacatalogus.nl). https://groningendatacatalogus.nl ER -